Model Selection

systematica.preprocessing.model_selection

get_aic_nb

get_aic_nb(
    arr: numpy.ndarray,
) ‑> float

Calculate the Akaike Information Criterion (AIC) from a log-likelihood array.

The AIC is defined as:

$$AIC = -2 * loglikelihood + 2 * k$$

where:

• log-likelihood is the sum of the log-likelihood values in the input array,
• k is the number of estimated parameters (assumed to be 1 in this implementation).

Parameters:

Name	Type	Default	Description
arr	tp.Array1d	--	A one-dimensional array containing log-likelihood values.

Returns:

Type	Description
float	The computed AIC value if the input contains valid data; otherwise, NaN.

get_bic_nb

get_bic_nb(
    arr: numpy.ndarray,
) ‑> float

Calculate Bayesian Information Criterion (BIC) from a log-likelihood array.

The BIC is defined as:

$$BIC = -2 * loglikelihood + k * log(n)$$

where:

• loglikelihood is the sum of the log-likelihood values in the input array,
• k is the number of estimated parameters (assumed to be 1 in this implementation),
• n is the number of observations (length of the input array).

Parameters:

Name	Type	Default	Description
arr	tp.Array1d	--	A 1-dimensional array of log-likelihood values.

Returns:

Type	Description
float	The BIC value if valid data is provided. Returns NaN if the input array is empty or contains no valid data.

get_aicc_nb

get_aicc_nb(
    arr: numpy.ndarray,
) ‑> float

Calculate Corrected Akaike Information Criterion (AICc) from a log-likelihood array.

The AICc is defined as:

$$AICc = AIC + 2k \frac{k + 1}{n - k - 1}$$

where:

• AIC is the Akaike Information Criterion,
• k is the number of estimated parameters (assumed to be 1 in this implementation),
• n is the number of observations (length of the input array).

Parameters:

Name	Type	Default	Description
arr	tp.Array1d	--	A 1-dimensional array of log-likelihood values.

Returns:

Type	Description
float	The AICc value if valid data is provided. Returns NaN if the input array is empty or contains no valid data.